The Effect of Proximal and Distal Training on Stroke Recovery: Stroke is the leading cause of impairment and disability. Although treatment of post stroke recovery in dedicated rehabilitation units positively influences outcome, no one treatment strategy has demonstrated superiority. We have been delivering targeted sensorimotor training for the shoulder and elbow with a planar robotic device, MIT-Manus. In randomized controlled trials, for both inpatients and persons with chronic impairment after stroke, persons treated with the robotic protocol demonstrated reduction in impairment in the exercised limb. Our results are in agreement with a prominent theme of current research into recovery from brain injury which posits that activity-dependent plasticity underlies neuro-recovery. If that is true, it suggests that significant recovery of motor function may be facilitated by "properly targeted" sensory-motor activity. By "properly targeted," we mean that with the under-emphasis or absence of sensory and motor stimulation of the central nervous system related either to a particular group of muscles and joints, or to particular tasks, the associated neural systems would be expected to (1) exhibit little plastic change, or (2) be pre-empted to subserve other functions. We have recently introduced to the clinic a complement to MIT-MANUS (shoulder-and-elbow robot): a wrist robot, which motivated the specific aims of this project: Specific Aim 1. Test whether task specific wrist robotic training improves motor performance among inpatients and persons with chronic impairment after stroke. Specific Aim 2. Test whether the order in which robot therapy is delivered influences outcomes (shoulder-and-elbow before wrist vs. wrist before shoulder-and-elbow). Specific Aim 3. Test whether there is generalization across different joints (shoulder & elbow vs wrist). Specific Aim 4. Test whether simultaneous targeted sensorimotor training of the shoulder, elbow, and wrist (integrated system) leads to better outcomes than isolated movement component training. Briefly we will invite persons with chronic impairment due to stroke to participate in a study that will train them first on wrist and then on the shoulder-and-elbow or vice-versa. A third group will be trained on the shoulder, elbow, and wrist simultaneously. Outcomes will be measured using standard instruments as well as robot-based measures.